# Channel-adaptive Cross-modal Generative Semantic Communication for Point Cloud Transmission

arXiv: https://arxiv.org/abs/2506.03211 · HTML: https://mlsanity.com/paper/2506.03211/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 30 references that are in the map.
- First posted on arXiv 3 Jun 2025.

## Builds on (most cited first, 19 of 19)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation](https://mlsanity.com/paper/1612.00593/) (2016)
- [PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space](https://mlsanity.com/paper/1706.02413/) (2017)
- [ShapeNet: An Information-Rich 3D Model Repository](https://mlsanity.com/paper/1512.03012/) (2015)
- [Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning](https://mlsanity.com/paper/1602.07261/) (2016)
- [Diffusion Probabilistic Models for 3D Point Cloud Generation](https://mlsanity.com/paper/2103.01458/) (2021)
- [Point-BERT: Pre-training 3D Point Cloud Transformers with Masked Point Modeling](https://mlsanity.com/paper/2111.14819/) (2021)
- [PC2: Projection-Conditioned Point Cloud Diffusion for Single-Image 3D Reconstruction](https://mlsanity.com/paper/2302.10668/) (2023)
- [PSTNet: Point Spatio-Temporal Convolution on Point Cloud Sequences](https://mlsanity.com/paper/2205.13713/) (2022)
- [Adaptive Graph Convolution for Point Cloud Analysis](https://mlsanity.com/paper/2108.08035/) (2021)
- [KeypointNet: A Large-scale 3D Keypoint Dataset Aggregated from Numerous Human Annotations](https://mlsanity.com/paper/2002.12687/) (2020)
- [Point Cloud Pre-training with Diffusion Models](https://mlsanity.com/paper/2311.14960/) (2023)
- [Back to 3D: Few-Shot 3D Keypoint Detection with Back-Projected 2D Features](https://mlsanity.com/paper/2311.18113/) (2023)
- [Digital-SC: Digital Semantic Communication with Adaptive Network Split and Learned Non-Linear Quantization](https://mlsanity.com/paper/2305.13553/) (2023)
- [Large Generative Model Assisted 3D Semantic Communication](https://mlsanity.com/paper/2403.05783/) (2024)
- [Semantic-aware Transmission for Robust Point Cloud Classification](https://mlsanity.com/paper/2306.13296/) (2023)
- [Deep joint source-channel coding for wireless point cloud transmission](https://mlsanity.com/paper/2408.04889/) (2024)
- [A Semantic Communication System for Real-time 3D Reconstruction Tasks](https://mlsanity.com/paper/2412.01191/) (2024)
- [Wireless 3D Point Cloud Delivery Using Deep Graph Neural Networks](https://mlsanity.com/paper/2006.09835/) (2020)

## Built on by (most cited first, 1 of 1)

- [Task-Oriented Wireless Transmission of 3D Point Clouds: Geometric Versus Semantic Robustness](https://mlsanity.com/paper/2603.13560/) (2026)

## Abstract (arXiv)

With the rapid development of autonomous driving and extended reality, efficient transmission of point clouds (PCs) has become increasingly important. In this context, we propose a novel channel-adaptive cross-modal generative semantic communication (SemCom) for PC transmission, called GenSeC-PC. GenSeC-PC employs a semantic encoder that fuses images and point clouds, where images serve as non-transmitted side information. Meanwhile, the decoder is built upon the backbone of PointDif. Such a cross-modal design not only ensures high compression efficiency but also delivers superior reconstruction performance compared to PointDif. Moreover, to ensure robust transmission and reduce system complexity, we design a streamlined and asymmetric channel-adaptive joint semantic-channel coding architecture, where only the encoder needs the feedback of average signal-to-noise ratio (SNR) and available bandwidth. In addition, rectified denoising diffusion implicit models is employed to accelerate the decoding process to the millisecond level, enabling real-time PC communication. Unlike existing methods, GenSeC-PC leverages generative priors to ensure reliable reconstruction even from noisy or incomplete source PCs. More importantly, it supports fully analog transmission, improving compression efficiency by eliminating the need for error-free side information transmission common in prior SemCom approaches. Simulation results confirm the effectiveness of cross-modal semantic extraction and dual-metric guided fine-tuning, highlighting the framework's robustness across diverse conditions, including low SNR, bandwidth limitations, varying numbers of 2D images, and previously unseen objects.